Texas Study Links Credit Scores to Race and Losses


AUSTIN, Texas January 05, 2005, By R.J. Lehmann, associate editor (BestWire) - Emboldening partisans on both sides of the credit-based insurance-scoring debate, a new study from the Texas Department of Insurance suggests consumer credit information may accurately predict the likelihood of homeowners and auto claims, while also finding that use of such information in underwriting disproportionately impacts black and Hispanic policyholders.

Using data culled from the credit-scoring models of six personal-lines insurer groups operating in Texas representing roughly 1.2 million personal auto policies and 800,000 homeowners policies the department presented the study, which was commissioned by a 2003 state law, to Texas Gov. Rick Perry and members of the state Legislature, whose 79th session begins later this month.

"We were charged with looking at the issue from a data perspective, and we took the time and effort to craft the best study we could come up with in the time frame," department spokesman Jerry Hagins said. "We're very pleased with the thoroughness of it. We think it will be really helpful in guiding the policy deliberations during our upcoming session."

The initial study which uses a "univariate" analysis that compares credit score to pure premium, in the case of auto, and loss ratio, in the case of homeowners finds that "there appears to be a strong relationship between credit score and insurance risk." In the case of the personal auto policies, as credit scores worsened, the researchers found that the average loss per vehicle increased, while a similar, though less definitive, trend also was found in declining loss ratios for better credit risks, the department reported.

Nonetheless, the department cautioned that credit-based scoring simply may reflect other characteristics, such as income, and it was less clear to what extent such other factors may or may not be even more predictive of loss experience. Toward that end, the department expects to complete its more comprehensive "multivariate" study by Jan. 31.

"It is necessary to evaluate if, and to what extent, credit scoring enables an insurer to more accurately predict losses," the researchers wrote. "Thoroughly analyzing this issue requires simultaneous analysis of the variables affecting likely claims experience."

Underscoring the need for a more extensive study was the department's finding that the state's blacks and Hispanics tend to be over-represented in the worst credit score categories and under-represented in the better credit score categories, the researchers noted. Using supplemental demographic data from the Texas Department of Public Safety, the researchers found that, as a whole, blacks had average credit scores that were 10% to 35% worse than the average scores for whites, while Hispanics had average scores that were 5% to 25% worse than those for whites. The average credit scores for Asian-Americans were about the same or slightly worse than those for whites, the researchers found.

To identify Hispanics, who are categorized as a separate race by DPS statistics, the department sorted policyholders according to a list of Hispanic surnames provided by the Texas Office of the Secretary of State, acknowledging that, in doing so, "some Hispanics were likely miscoded as white, while some whites were likely miscoded as Hispanic."

Credit scores are used to determine rates or rating tiers by roughly 42% of insurers writing homeowners in the state, and about 55% of those writing personal auto, the study found. In the personal auto market, 90% of policyholders experienced a rate swing as a result of an insurer's use of credit information, with 10% experiencing increases or decreases of 40% or more. In the homeowners market, 90% of policyholders experienced either no rate swing or one of less than 30%.

Immediate reaction to the study was mixed, as both supporters and opponents of credit-based scoring were quick to point to the department's findings as supporting their position.

"The Legislature was concerned about credit-scoring discriminating on the basis of race, which is why they asked the department to do the study," said Birny Birnbaum, executive director of the Austin-based Center for Economic Justice. "They now have the results of the study. It does discriminate on the basis of race. So now that they've found the answer, it seems pretty obvious what they need to do ban credit-scoring. If you're not going to allow somebody to discriminate on the basis of race directly, then why would you allow them to use a proxy to discriminate on the basis of race indirectly?"

In contrast, the Insurance Council of Texas pointed to the study as vindication of the industry's oft-stated position that insurance scoring is an objective and "colorblind" process that has proven itself an accurate predictor of insurance losses.

"The use of credit scores is an everyday fact of life," council spokesman Mark Hanna said. "They are used in obtaining jobs, finding an apartment, purchasing a home and buying nearly every type of major appliance. Credit scores are an underwriting factor that an individual can control and improve. By paying bills on time and managing their finances, policyholders can see a marked improvement in their credit score, which can lead to better insurance rates."

Striking a somewhat more cautious tone, the Property Casualty Insurers Association of America characterized the study as a "preliminary review" and urged lawmakers to reserve judgment until a more detailed analysis was available.

"Studies that attempt to link credit information with race or income are extremely complex," PCI said in a statement. "While several states such as Alaska, Missouri and Washington have attempted to conduct this type of study, each one has suffered from critical flaws in the methodology and failed to adequately consider loss histories, which rendered the results invalid."

But pointing to the findings of previous studies that were similar to those of the Texas study, Birnbaum said the links between credit scores and race now are so well-established as to make debating the point futile.

"Even if you control for income, you will find that credit-scoring discriminates on the basis of race," Birnbaum said. "That's consistent with research that I've done in the past on insurance availability. We've done analyses showing that your likelihood of being rejected by the standard and preferred auto market is four times as great if you come from a poor, minority neighborhood as if you come from an affluent, white neighborhood."